On Sat, Nov 08, 2025 at 01:05:55AM +0000, Tian, Kevin wrote: > > From: Jason Gunthorpe <[email protected]> > > Sent: Saturday, November 8, 2025 8:48 AM > > > > On Fri, Nov 07, 2025 at 03:10:33AM +0000, Tian, Kevin wrote: > > > > To me, it looks like something generic, that will have impact on any > > > > device specific driver variant. > > > > What am I missing? > > > > > > > > I wonder if drivers that don't implement the deferred reset trick were > > > > ever executed with lockdep enabled. > > > > > > > > > > @Jason, @Yishai, @Shameer, @Giovanni, @Brett: > > > > > > Sounds it's a right thing to pull back the deferred reset trick into > > > every driver. anything overlooked? > > > > It does seem like we should probably do something in the core code to > > help this and remove the duplication. > > from backport p.o.v. it might be easier to first fix each driver > independently then remove the duplication in upstream?
If it hasn't bothered anyone yet I wouldn't stress about backporting.. Maybe those drivers do work for some unknown reason? Plus it is *really* hard to actually hit this deadlock.. Jason
